TWI790414B - Feature recognition method and system for human body - Google Patents

Abstract

A feature recognition method and system for human body, capable of identifying a feature target and a human body contour in an input picture or video, and distinguishing the recognized human body contour into several body parts regions, Then first define the coordinates of the feature target and the coordinates of the body part area in the human body contour, and then match the feature coordinates with the coordinates of the human body part area to determine whether the feature target is located in the body contour within any or multiple body part regions.

Description

Feature recognition method and system for human body

The present invention relates to a feature recognition method and system for human body, especially a feature recognition method and system capable of judging whether a feature is located in any or any multiple body parts of the human body.

At present, the common image recognition is mostly for the recognition of a specific area (such as facial recognition), but for some special needs, it is necessary to recognize the people and objects on the screen, but most of the recognition methods are only the target object Marking and frame selection, but it is very difficult to further identify the features and locations on the human body in a more subtle way.

In terms of accidental burn wounds or scald wounds, when medical staff take care of the wound, they often observe with the naked eye, and record and describe the wound condition through text. However, identifying the wound with the naked eye is very likely because Different medical staff have doubts about misjudgment. Therefore, if the identification method and system of this case can be used, in addition to accurately recording the position of physical features, it will also allow medical staff to remotely understand the status of the injured patient. Therefore, The present invention should be an optimal solution.

The present invention is used in the feature recognition method and system of the human body, the steps of which are as follows: (1) In an input picture or video, identify a characteristic target and a human body contour, and divide the recognized human body contour into several body parts regions; (2) define the coordinates of the characteristic target and the coordinates of the body part area within the human body outline; (3) match the coordinates of the feature target with the coordinates of the body part area of the human body to determine whether the feature target is located in any or any body parts of the human body outline within the site area.

More specifically, several human body coordinate points can be used to distribute in the human body contour, and the body part area in the human body contour is distinguished by the area formed by the several human body coordinate points.

More specifically, the center point of the characteristic target can be found out, and the coordinates of the center point are matched with the human body coordinate points distributed in the outline of the human body to find out the human body coordinates with the same coordinates as the center point. point or the human body coordinate point closest to the center point, and judge whether the characteristic target is located in which one or which multiple body part regions according to the range area to which the matching belongs.

More specifically, according to the distance between the center point and the human body coordinate point, the order of distance between the center point and different human body coordinate points can be sorted out, and according to the sorting result and the range area to which the different human body coordinate points belong , to provide the possibility judgment information that the feature target is located in different body regions.

More specifically, the feature coordinate range of the feature target shape can be found, and the human body coordinate points distributed in the human body contour are matched with the feature coordinate range to find out which human body coordinate points Whether it is within the coordinate range of the feature, and according to the number of coordinate points of the human body within the coordinate range of the feature and the range area to which it belongs, determine which one or several body part areas have the feature target.

More specifically, the said can be based on a certain one distributed within the coordinate range of the feature The maximum number of human body coordinate points in the range area is used to judge which body part area has the characteristic target.

More specifically, according to the number of human body coordinate points distributed within the coordinate range of the feature and the range area to which it belongs, the distribution size of different range areas within the coordinate range of the feature can be screened out, and based on the screening results, the Judging the distribution relationship of the characteristic target located in different body parts and providing the possibility judgment information of the characteristic target located in different body regions.

A feature recognition system for the human body, which is set on an electronic device, and the electronic device has at least one processor and at least one computer-readable recording medium, and the computer-readable recording medium stores one or more pictures file or video file, wherein the computer-readable recording medium further stores at least one application unit capable of processing feature recognition, and when the application unit is executed by the processors, the electronic device is caused to perform the following Program: In the activated application unit, the input picture or video can be used to identify a characteristic target and a human body contour, wherein the human body contour is divided into several body parts, and then the characteristic target and the human body contour are defined and then match the feature coordinates with the coordinates of the human body part area to determine whether the position of the feature object is located in any or any body part area of the human body outline.

More specifically, the application unit includes: an input module for inputting one or more pictures or videos; an object detection module connected with the input module for the The image or video input by the input module is used for object detection to identify one or more characteristic targets and their coordinates; a human body detection module is connected with the input module to use the input module The input picture or video is used for human detection to identify one or more human body contours, and the recognized human body contours are divided into several body parts and their coordinates; and a position matching module, which is connected with the human body The object detection module is connected with the human body detection module to determine whether the characteristic target is located on the human body within any or any number of body part regions of the contour.

A feature recognition method for a human body, the steps of which are: (1) identify a feature target and a human body outline in an input picture or video; (2) define the coordinates of the feature target and the human body (3) matching the coordinates of the characteristic target with the coordinates in the human body contour to determine whether the characteristic target is located in the human body contour.

More specifically, several human body coordinate points can be distributed within the contour of the human body.

More specifically, the center point of the characteristic target can be found out, and the coordinates of the center point are matched with the human body coordinate points distributed in the outline of the human body to find out the human body coordinates with the same coordinates as the center point. point, to determine whether the characteristic target is located in the outline of the human body, and when extending to a multi-person environment, it can find out which person the characteristic target is on.

More specifically, the feature coordinate range of the feature target shape can be found, and the human body coordinate points distributed in the human body contour are matched with the feature coordinate range to find out which human body coordinate points Whether it is within the coordinate range of the characteristic object is used to determine whether the characteristic object is located within the outline of the human body, and when extended to a multi-person environment, it can find out which person the characteristic object is on.

A feature recognition system for the human body, which is set on an electronic device, and the electronic device has at least one processor and at least one computer-readable recording medium, and the computer-readable recording medium stores one or more pictures file or video file, wherein the computer-readable recording medium further stores at least one application unit capable of processing feature recognition, and when the application unit is executed by the processors, the electronic device is caused to perform the following Program: In the activated application unit, the input picture or video can be recognized to identify a characteristic target and a human body outline, and then define the characteristic coordinates of the target and the human body contour, and then match the feature coordinates with the coordinates of the human body contour to determine whether the position of the characteristic target is within the human body contour.

More specifically, the application unit includes: an input module for inputting one or more pictures or videos; an object detection module connected with the input module for the The image or video input by the input module is used for object detection to identify one or more characteristic targets and their coordinates; a human body detection module is connected with the input module to use the input module Human body detection is performed on input pictures or videos to identify one or more human body contours and their coordinates; and a position matching module is connected with the object detection module and the human body detection module for use in In order to judge whether the characteristic object is located in the contour of the human body.

1: Electronic equipment

11: Processor

12: Computer-readable recording media

121: Media file storage area

122: Application unit

1221: input module

1222: Object detection module

1223:Human detection module

1224: Position matching module

2: Picture

21: Feature Target

22: Human silhouette

3: Picture

31: Feature Target

32: Human silhouette

[Fig. 1] is a schematic flow chart of the feature recognition method and system for the human body according to the present invention.

[Fig. 2A] is a schematic diagram of the system architecture of the feature recognition method and system for the human body according to the present invention.

[Fig. 2B] is a schematic diagram of the structure of the electronic equipment used in the feature recognition method of the human body and its system according to the present invention.

[Fig. 2C] is a schematic diagram of the application unit architecture of the feature recognition method for human body and its system according to the present invention.

[Fig. 3A] is a schematic diagram of the first implementation of the feature recognition method and system for the human body according to the present invention.

[Fig. 3B] is a schematic diagram of the first implementation of the feature recognition method and system for the human body according to the present invention.

[Fig. 3C] is a schematic diagram of the first implementation of the feature recognition method and system for the human body according to the present invention.

[Fig. 4A] is a schematic diagram of the second implementation of the feature recognition method and system for the human body according to the present invention.

[Figure 4B] is a schematic diagram of the second implementation of the feature recognition method and system for the human body according to the present invention.

[Fig. 4C] is a schematic diagram of the second implementation of the feature recognition method and system for the human body according to the present invention.

[Fig. 5] is another schematic flow chart of the feature recognition method and system for the human body according to the present invention.

Other technical contents, features and effects of the present invention will be clearly presented in the following detailed description of preferred embodiments with reference to the drawings.

Please refer to Figure 1, which is a schematic flow chart of the feature recognition method and system for the human body according to the present invention. As can be seen from the figure, the steps are: (1) Identify a feature target in an input picture or video and a human body contour, and distinguish the recognized human body contour into several body part regions 101; (2) define the coordinates 102 of the characteristic target and the body part region in the human body contour; (3) divide The coordinates of the characteristic object are matched with the coordinates of the body part area of the human body to determine whether the characteristic object is located in any one or more body part areas of the human body outline 103 .

Please refer to Figures 2A, 2B and 2C, which are schematic diagrams of the system architecture, electronic equipment architecture, and application unit architecture of the feature recognition method for the human body and its system according to the present invention. It is known that the system is set in an electronic device 1 (the electronic device can be a server device or a computer device or an integrated computer device), and the electronic device 1 has at least one processor 11 and at least one computer can Read the recording medium 12, wherein the computer-readable recording medium 12 has a media file storage area 121 and an application unit 122, and the media file storage area 121 has one or more picture files or video files; When the application unit 122 is executed by the processors 11, the electronic device 1 is caused to perform the following procedures: in the activated application unit 122, the input picture or video can be recognized to identify a characteristic target and a human body outline , where the human body contour system is divided into several body part regions, and then define the coordinates of the feature target and the body part region in the human body contour, and then match the feature coordinates with the coordinates of the human body part region, to It is judged whether the feature target is located in any or any body part area of the human body contour.

And the application unit 122 includes: (1) an input module 1221 for inputting one or more pictures or videos; (2) an object detection module 1222 connected with the input module 1221 , for object detection of pictures or videos input by the input module 1221, to identify one or more characteristic targets; (3) a human body detection module 1223, which is connected with the input module 1221 , for performing human body detection on the pictures or videos input by the input module 1221, to identify one or more human body contours, and to distinguish the identified human body contours into several body parts regions; and (4) A position matching module 1224 is connected with the object detection module 1222 and the human body detection module 1223 to determine whether the feature object is located in any or any body part area of the human body contour.

The "whether it is located" mentioned above includes the following three judgments: (1) It can be judged that "the characteristic object is not located in any or any body part area of the human body contour"; (2) it can be judged that "the characteristic object has any or any body part area located in the human body contour"; (3 ) can judge "which one or more body parts of the body contour the characteristic target is located in"; (4) the above three judgment modes can be judged in a single mode, or any two or three judgments can be carried out simultaneously.

And when there is already a characteristic target on the human body (but the technology of the present invention can also be applied to judge when the characteristic target is not located on the human body), when it is necessary to perform positioning recognition, as shown in Figure 3A, in Figure 2 There is a human body, the human body has a wound/scar, and the judgment steps for judging whether the wound/scar is located in which part of the human body are as follows: (1) In Picture 2, as shown in Figure 3B, identify a feature The target 21 and a human body contour 22 come out, and can be distributed in the human body contour with several human body coordinate points, and the body part area system in the human body contour is distinguished by the different range regions formed by several human body coordinate points; 2) Afterwards, as shown in Fig. 3C, the center point of the characteristic target 21 can be found (according to the whole picture, find out the center point of the characteristic target 21, and define the center point based on the whole picture The coordinates of the center point are (x ₁ , y ₁ )), and the coordinates (x ₁ , y ₁ ) of the center point are matched with the coordinate points of the human body distributed in the outline of the human body (the center point system can be the geometric center point, the center of weight Point or relative center point, such as the exact center point of the bounding box, the geometric center point of the feature, the average (or median) of all x-coordinates (or median) and the average (or median) of y-coordinates calculated (x, y ), or other unrestricted center point definition methods); (3) and the following three modes can be available in the matching process:

(a) The first mode can filter out a human body coordinate point closest to the center point based on the shortest distance between the center point and the human body coordinate point in a certain area, so as to determine the characteristic target Whether it is located in which or which multiple body parts areas (as can be seen from the 3C figure, the center point is the closest position or the human body coordinate point in the range area that is the same as the right arm position, so it is judged whether the characteristic target 21 Located on the right arm position; if there are two human body coordinate points (at the right arm position and at the body torso position) and the center point are all the same distance, then it can be determined whether the feature target 21 is located at the right arm position or the body torso position superior).

(b) The second mode can sort out the distance order between the center point and different human coordinate points according to the distance between the center point and the human body coordinate point, and according to the sorting result and the range to which different human body coordinate points belong area, to determine the far-short relationship between the feature target and different body parts, so as to provide the possibility judgment information of the feature target in different body areas (taking Figure 3C as an example, the position of the right arm can be listed as the first one close, while the body torso position is the second close, and the head position is the third close order of distance).

(c) The third mode can also be screened according to the threshold value, that is, within a certain range of the number of human body coordinate points, it is screened out which ranges are close to the center point.

And when it is necessary to carry out another kind of positioning recognition, as shown in FIG. 4A, there is a human body in FIG. 3 with a wound/scar on the body, and the judgment steps for judging whether the wound/scar is located on which part of the human body are as follows: (1) In picture 3, as shown in Figure 4B, a characteristic target 31 and a human body contour 32 are identified, and can be distributed in the human body contour with several human body coordinate points, and the human body contours The body part area is distinguished by the different range areas formed by several human body coordinate points; (2) afterwards, as shown in Figure 4C, the feature object coordinate range (( x ₁ ,y ₁ )(x ₂ ,y ₂ )(x ₃ ,y ₃ )(x ₄ ,y ₄ )), and match the human body coordinate points distributed in the human body contour with the feature coordinate range; And (3) can have the following two modes in the matching process:

(a) The first mode can be used to screen which one or more body part areas have characteristic targets according to the maximum number of human body coordinate points distributed in a certain area within the feature coordinate range; ( a1) If there is a group of human body coordinate points (all in the right arm position) falling within the coordinate range of the feature target, it is judged whether the feature target is located in the right arm position; (a2) if there are two sets of human body coordinate points (in the right arm position position and at the body trunk position) fall within the characteristic target coordinate range, as shown in Figure 4C, the number of human body coordinate points of the two groups is the same or close, so it can be judged whether the characteristic target 21 is located at the right arm position and the body position between torso positions.

(b) The second mode can sort out the order of the distribution of different range areas within the coordinate range of the feature according to the number of human body coordinate points distributed within the coordinate range of the feature and the range area to which they belong, and according to the sorting result , to determine the distribution size relationship of the feature target located in different body parts, and provide the possibility judgment information of the feature target located in different body areas; if the number of human coordinate points of the right arm position, body trunk position and head position ( 40, 25, 10) are inconsistent, it can be listed that the position of the right arm is the first in the number of coordinate points of the human body, the position of the torso is the second, and the position of the head is the third. order).

In addition, in this case, it is not necessary to separate the body parts, but only to determine whether the feature object is located in the contour of the human body, as shown in Figure 5, the processing steps are as follows: (1) Identify in an input image or video Go out a characteristic target and a human body contour and come out 501; (2) define the coordinate 502 in this characteristic target and this human body contour; (3) match the coordinate of this characteristic target and this human body contour, with It is judged whether the feature object is located within the human body contour 503 .

And the recognition mode is as follows:

(1) It is possible to find out the center point of the characteristic target, and match the coordinates of the center point with the human body coordinate points distributed in the human body contour to find out the human body coordinate points with the same coordinates as the center point position, and according to Match the range area it belongs to to determine the location of the feature object.

(2) It is possible to find out the feature coordinate range of the feature target shape, and match the human body coordinate points distributed in the human body outline with the feature coordinate range to find out which human body coordinate points are located in the feature within the range of object coordinates.

For the identification purpose, the human body detection module 1223 only needs to identify one or more human body contours and their coordinates, and then judge whether the characteristic object is located in the human body contour through the position matching module 1224 .

The feature recognition method and system for the human body provided by the present invention, when compared with other conventional technologies, have the following advantages:

(1) The present invention can be used to identify features and locations on the human body. When applied to the medical field, in addition to being able to clearly search and record the location of wounds on the body, it can also allow medical staff to remotely Know the status of the injured, so it will be very useful for the medical field help.

(2) The present invention can first identify whether there are clothes and mark the range of the clothes, and then when identifying the characteristic target, it can exclude the range marked by the clothes to identify whether the characteristic target is located in the range marked by the clothes Inside, to avoid misjudgment caused by coats or other clothes.

The present invention has been disclosed above through the above-mentioned embodiments, but it is not intended to limit the present invention. Anyone who is familiar with this technical field and has common knowledge can understand the foregoing technical characteristics and embodiments of the present invention without departing from the present invention. Within the spirit and scope, some changes and modifications can be made, so the patent protection scope of the present invention must be defined by the claims attached to this specification.

Claims (13)

A feature recognition method for a human body, the steps of which are: in an input picture or video, an application unit identifies a feature target and a human body contour, and distinguishes the recognized human body contour into several bodies Part area, wherein the application unit can be distributed in the body contour with several human body coordinate points, and the body part area in the human body contour is distinguished by the range area formed by several human body coordinate points; the application unit defines Find the coordinates of the characteristic target and the coordinates of the body part area in the human body contour; the application unit matches the coordinates of the characteristic target with the coordinates of the human body part region to determine whether the characteristic target is located in the human body contour within any or multiple body part regions. The feature recognition method for the human body as described in Claim 1, wherein the application unit can find out the center point of the feature object, and match the coordinates of the center point with the human body coordinate points distributed in the human body contour, so as to Find the human body coordinate point with the same coordinates as the center point or the human body coordinate point closest to the center point, and judge whether the feature target is located in which one or more body parts according to the matching range area within the area. The feature recognition method for human body as described in claim 2, wherein the application unit can sort out the order of distance between the center point and different human body coordinate points according to the distance between the center point and the human body coordinate point, and Based on the sorting results and the range areas to which different human body coordinate points belong, the possibility judgment information of the feature target located in different body areas is provided. The feature recognition method for the human body as described in Claim 1, wherein the application unit can find out the feature coordinate range of the feature target shape, and compare the human body coordinate points distributed in the human body outline with the feature coordinate range Match to find out which human coordinate points are located in the feature within the object coordinate range, and according to the number of human body coordinate points within the feature coordinate range and the range area to which it belongs, it is judged which one or which body part areas have the feature target. The feature recognition method for the human body as described in claim 4, wherein the application unit can judge which body part area is in the area according to the maximum number of human body coordinate points distributed in a certain area within the feature coordinate range with this feature target. The feature recognition method for the human body as described in claim 4, wherein the application unit can filter out different range areas in the feature coordinates according to the number of human body coordinate points distributed within the feature coordinate range and the range area to which they belong The size of the distribution within the range, and according to the screening results, determine the distribution relationship of the characteristic target located in different body parts and provide information on the possibility of the characteristic target located in different body regions. A feature recognition system for the human body, which is set on an electronic device, and the electronic device has at least one processor and at least one computer-readable recording medium, and the computer-readable recording medium stores one or more pictures file or video file, wherein the computer-readable recording medium further stores at least one application unit capable of processing feature identification, and when the application unit is executed by the processors, the electronic device causes the following Program: In the activated application unit, the input picture or video can be recognized to identify a characteristic target and a human body contour, wherein the human body contour system is divided into several body parts areas, which can be distributed by several human body coordinate points In the human body contour, and the body part area in the human body contour is distinguished by the range area formed by several human body coordinate points, defining the coordinates of the feature target and the body part area in the human body contour, and then The coordinates of the feature are matched with the coordinates of the body part area of the human body to determine whether the position of the feature object is located in any one or more body part areas of the human body outline. The feature recognition system for the human body as described in Claim 7, wherein the application unit includes: an input module for inputting one or more pictures or videos; an object detection module for the The input module is connected to perform object detection on pictures or videos input by the input module, so as to identify one or more characteristic targets and their coordinates; a human body detection module is connected to the input module Connected to perform human body detection on pictures or videos input by the input module, to identify one or more human body contours, and to distinguish the recognized human body contours into several body parts and their coordinates; And a position matching module, which is connected with the object detection module and the human body detection module, to determine whether the characteristic target is located in any or any body part area of the human body outline. A feature recognition method for a human body, the steps of which are: in an input picture or video, an application unit identifies a feature target and a human body outline, wherein several human body coordinate points can be distributed on the human body inside the contour; the application unit defines the coordinates of the characteristic target and the coordinates in the human body contour; the application unit matches the coordinates of the characteristic target with the coordinates in the human body contour to determine whether the characteristic target is located in the human body inside the outline. The feature recognition method for the human body as described in Claim 9, wherein the application unit can find out the center point of the feature object, and match the coordinates of the center point with the human body coordinate points distributed in the human body contour, so as to Find the coordinate point of the human body with the same coordinates as the center point, so as to judge whether the feature target is located in the human body contour. The feature recognition method for the human body as described in Claim 9, wherein the application unit can find out the feature coordinate range of the feature target shape, and distribute the human body within the human body contour The coordinate points are matched with the coordinate range of the feature to find out which coordinate points of the human body are within the coordinate range of the feature, so as to determine whether the feature target is within the outline of the human body. A feature recognition system for the human body, which is set on an electronic device, and the electronic device has at least one processor and at least one computer-readable recording medium, and the computer-readable recording medium stores one or more pictures file or video file, wherein the computer-readable recording medium further stores at least one application unit capable of processing feature identification, and when the application unit is executed by the processors, the electronic device causes the following Program: In the activated application unit, the input picture or video can be used to identify a characteristic target and a human body contour, in which several human body coordinate points can be distributed in the human body contour, and then the feature can be defined coordinates of the target and the human body contour, and then match the feature coordinates with the coordinates of the human body contour to determine whether the position of the characteristic target is within the human body contour. The feature recognition system for the human body as described in claim 12, wherein the application unit includes: an input module for inputting one or more pictures or videos; an object detection module for the The input module is connected to perform object detection on pictures or videos input by the input module, so as to identify one or more characteristic targets and their coordinates; a human body detection module is connected to the input module connected to detect the human body on the pictures or videos input by the input module, so as to identify one or more human body contours and their coordinates; and a position matching module, which is connected with the object detection module and The human body detection module is connected to determine whether the characteristic target is located within the human body contour.

Images